SI-traceable quantification of sulphur in copper metal and its alloys by ICP-IDMS

Abstract

Previously applied methods for the quantification of sulphur in copper and other pure metals revealed a lack of SI-traceability and additionally showed inconsistent results, when different methods were compared. Therefore, a reference procedure is required which allows SI-traceable values accompanied by a sound uncertainty budget. In this study a procedure was developed for the quantification of total sulphur in copper at low concentration levels using inductively coupled plasma-isotope dilution mass spectrometry (ICP-IDMS). The major part of the copper matrix was separated by adding ammonia which forms a complex with the copper while releasing the sulphur followed by chromatographic separation using a weak cation resin. After that the sulphur fraction was further purified by chromatographic means using first an anion exchange method and second a chelating resin. The developed procedure shows high performance, especially concerning high efficiency in matrix removal (>99.999%) while keeping the recovery of sulphur above 80%. Procedure blanks are in the order of 3–53 ng resulting in LOD and LOQ values of 0.2 μg g⁻¹ and 0.54 μg g⁻¹, respectively. The procedure is sufficient to facilitate value assignment of the total sulphur mass fraction in reference materials. Additionally, relative measurement uncertainties were calculated to be below 1% and the measurement results were traceable to the SI. The procedure reported in this study is a new reference procedure for sulphur measurement in copper, being fit for two major purposes, certification of reference materials and assignment of reference values for inter-laboratory comparison.

---

1. Introduction

Copper is an essential metal for different organisms and it has been one of the most useful metals known to man. It is the third most important metal in industrial applications, and it is widely used in unalloyed and alloyed conditions. The copper industry in Europe has an estimated turnover of about €45 billion, and copper prices have increased by a factor of 4 in the last decade. Mainly copper is used in electricity and energy, building construction, engineering and transportation.¹ The world's copper supply comes from two main sources: mining/refining and recycling. Copper ores are normally associated with sulphur, and copper can be extracted from chalcocite (Cu₂S, copper sulphide), chalcopyrite (CuFeS₂, copper iron sulphide) and cuprite (Cu₂O, copper oxide). Although there are more than 400 copper alloys in use, the applications mainly focus on copper–zinc alloys (brass), copper–tin alloys (bronze), copper–nickel alloys, copper–beryllium alloys and pure copper (unalloyed), which is mainly used in electrical engineering.

Keeping up the quality of copper and copper alloys in technology requires specific reference materials. In general, copper reference materials can be divided into two major classes, materials which are characterized by their copper mass fraction or purity (main element, kg kg⁻¹) and materials which are characterized by impurities (other elements). Sulphur (S) is one of the major impurities in copper which directly influences its chemical, physical, and mechanical properties such as the colour, hardness, tensile strength, and heat treatment of copper.^2,3 Therefore, the determination of the sulphur content in copper is necessary for many technological applications.

Available copper reference materials (RMs) providing data for the sulphur mass fraction are compiled in Table 1. Roughly half of the listed materials are certified for their sulphur mass fractions; the other half of the reference materials only provide additional/information values for the sulphur mass fraction. A large part of the materials being certified for sulphur show relative measurement uncertainties of 7–30%, whereas all other materials only provide relative standard deviations of the inter-laboratory comparison or no uncertainty data at all. The reviewed information emphasizes the lack of reference procedures, which can provide sufficiently small measurement uncertainties and which are suitable as reference procedures, especially for the certification of reference materials.

Table 1 Review of copper certified reference materials (CRMs) with reference values for the total sulphur mass fraction (for more information see ref. 5)

Material no.	Description	Copper mass fraction in, kg kg^−1	Sulphur mass fraction	CV/AI^a	U rel in^b, %	Comment	Producer, country	Year of issue	Available
a CV = certified value, AI = additional information for sulphur. b Relative measurement uncertainty.
BAM-Y001	Pure copper	99.9970 ± 0.001	5.4 ± 1.6 μg g^−1	CV	29.6	Certified by spectrophotometry	BAM, DE	2004	✓
BAM-M376a	Pure copper	—	133 ± 19 μg g^−1	AI	14.3	The uncertainty larger than expected or in case of possible inhomogeneities	BAM, DE	2016	✓
BAM-M385	Pure copper	—	31.2 ± 1.5 μg g^−1	CV	4.8	Certified by inter-laboratory comparison	BAM, DE	2003	✗
BAM-227	Copper chips	85.57 ± 0.03	0.122% ± 0.010% (2SD)	—	8.2	Certified by photometry	BAM, DE	1979	✗
BAM-228	Copper chips	85.34 ± 0.03	0.036% ± 0.004% (2SD)	—	11.2	Certified by coulometry and photometry	BAM, DE	1979	✓
NIST SRM399	Unalloyed copper – Cu VI	99.79 ± 0.01	(10) μg g^−1	AI	—	—	NIST, US	1993	✓
NIST SRM400	Unalloyed copper – Cu VII	99.70 ± 0.02	(9) μg g^−1	AI	—	—	NIST, US	1986	✓
NIST SRM457	Unalloyed copper IV (solid)	99.97 ± 0.18	(4 ± 1) μg g^−1	AI	25.0	Certified by inter-laboratory comparison including NIST	NIST, US	2013	✓
NIST SRM494	Unalloyed copper – Cu I	99.91 ± 0.01	(15 ± 3) μg g^−1	CV	20.0	Certified by inter-laboratory comparison	NIST, US	1986	✓
NIST SRM495	Unalloyed copper – Cu II	99.94 ± 0.01	(13 ± 2) μg g^−1	CV	15.4	Certified by inter-laboratory comparison	NIST, US	1987	✓
NIST SRM498	Unalloyed copper V (solid)	99.98 ± 0.01	(11) μg g^−1	AI	—	—	NIST, US	1993	✓
NIST SRM885	Refined copper	—	(18 ± 3) μg g^−1	CV	16.7	—	NIST, US	1991	✓
NIST SRM1034	Unalloyed copper	99.96 not certified	(2.8 ± 0.2) μg g^−1	CV	7.1	Certified by ID-TIMS technique	NIST, US	1982	✓
BCR-017-B	Pure copper	—	(10.4 ± 0.6) μg g^−1	CV	5.8	Certified by inter-laboratory comparison	IRMM, BE	1989	✓
IARM 278A	Tellurium copper	99.5	(20 ± 3) μg g^−1	CV	15	Certified by inter-laboratory comparison	ARMI, UK	2007	✓
39X 17866	Residuals in copper (CHILL-CAST)	—	(520 ± 30) μg g^−1	AI	5.8	—	MBH Analytical Ltd, UK	2014	✓
39X 17871	Residuals in copper (CHILL-CAST)	—	(52 ± 4) μg g^−1	AI	7.4	—	MBH Analytical Ltd, UK	2013	✓

---

Sayi et al.⁴ reviewed the different techniques available to determine the total sulphur mass fraction in inorganic compounds such as gravimetry, titrimetry, spectrophotometry, combustion, and mass spectrometry. The techniques were presented depending on the measurement range for the sulphur mass fraction and the corresponding precision in various sample types. The obtained precision of those techniques is in the range of 0.2–10% relative standard deviation. One of the most powerful techniques for the determination of sulphur in metal samples is glow discharge mass spectrometry (GDMS). Major advantages of this technique are a short analysis time and a simple sample preparation process, both making it fit for routine analysis. The disadvantages of GDMS, however, are relatively high measurement uncertainties of 20% and more and the requirement of matrix matched standards for obtaining reliable results. Classical methods such as carrier gas hot extraction/combustion analysis are used for sulphur analysis by converting the sulphur in the sample into sulphur dioxide in a high purity oxygen atmosphere, followed by the measurement of the sulphur dioxide by using thermal conductivity or infrared absorption spectrometry. This technique is quite fast but the precision is in the range of 2–10% (expressed in RSD) which is not fit for our purposes.⁴

In recent years, inductively coupled plasma-mass spectrometry (ICP-MS) has been investigated for the quantification of sulphur in low level materials. Martinez-Sierra et al. clearly reviewed the technical problems of sulphur analysis by ICP-MS such as required mass resolution and potential interference on the basis of various publications. Most of the applications, however, are focused on organic samples such as fuel, protein, and pharmaceuticals.⁶

A major challenge for the quantification of sulphur in copper (alloyed/unalloyed) by ICP-MS is the copper matrix itself, causing matrix effects and making extensive cleaning of cones and extraction lenses necessary after measurements. Matschat et al. investigated the analysis of high-purity metals (including copper) by high resolution ICP-MS.⁷ They found that the copper matrix shows strong matrix effects on the sensitivity, resulting from Cu deposition on the cones. The relative decrease in the sensitivity amounts to about 70% when aspirating a 5000 mg L⁻¹ copper solution.⁷ Moreover, the compared analytical methods for analysing impurities in pure copper, but the number of reported datasets were small while the standard deviations were large.⁸ Most commonly matrix effects in ICP-MS are being reduced by sufficient dilution, often with dilution factors of 10 000 and higher. In the case of sulphur analysis in copper such high dilution factors are ineffective, because as a consequence sulphur is diluted to the medium to low ng g⁻¹ range making measurements with sufficiently low measurement uncertainties (<10% relative) impossible.

The application of isotope dilution mass spectrometry (IDMS) can overcome some of these limitations, as it facilitates the use of matrix separation techniques. Since sample loss will not affect the accuracy of the results once equilibration between the sample and the spike is established. Additionally, IDMS enables the smallest measurement uncertainties while providing traceability to the international system of units (SI).^9,10 In combination with thermal ionization mass spectrometry (TIMS) IDMS has been applied to determine sulphur in metal samples^11,12 and fossil fuel samples.¹³ The sample preparation of ID-TIMS consists of two major steps; the sample digestion and the formation of As₂S₃. During digestion sulphur is oxidized to sulphate. This sulphate then has to be reduced to H₂S, transferred and finally converted into As₂S₃, which is laborious and time-consuming. Moreover, the relatively high blank values and their uncertainties strongly contribute to the combined uncertainties of the sulphur mass fraction.¹³

In this study, the IDMS concept is combined with ICP-MS analysis and a new approach for sulphur–copper separation is developed. The sample preparation including sample digestion, matrix removal by complexation and ion exchange chromatography, as well as critical points will be presented in detail. The aim of this research is to develop a reliable measurement procedure which enables low measurement uncertainties and SI traceability and which can be applied to the certification of reference materials, the assignment of reference values and the calibration of other analytical procedures. The target for the relative measurement uncertainty in this study is set to below 2%. In order to demonstrate the SI traceability, the unbroken chain of calibrations will be established.

2. Experimental

2.1 Materials and reagents

The determination of sulphur especially at the low μg g⁻¹ level requires clean working conditions and specialized sample handling equipment to keep the blank contribution and contamination risks small. All samples were prepared in a class 10 clean room (Fed STD 209E). All plastic laboratory equipment was soaked in 10% HNO₃ for at least 60 hours whereas PFA beakers were cleaned with a device using nitric acid vapours for cleaning and subsequent soaking in Milli-Q water for at least 12 h. Thereafter the laboratory equipment was dried by air flow in the cleanroom cabinets.

To keep the sulphur blank as low as possible all reagents were used in the highest available purity. Nitric acid, which was used for sample digestion and sulphur–matrix separation, was purified by a two-stage sub-boiling procedure. Ammonia solution (Suprapur®) and hydrogen peroxide (Ultrapur®) were obtained from Merck KgaA (Darmstadt, Germany).

As the primary assay, used as the back-spike in IDMS, the sulphur standard solution NIST SRM3154 was employed. The enriched isotope ³⁴S was obtained from Trace Sciences International Inc. (Delaware USA) in the form of solid sulphur, with a nominal enrichment of 99.8%. The ³⁴S was dissolved in HNO₃ to prepare the ³⁴S enriched spike solution, which then was characterized using back-spike solutions. Sodium sulphide (Na₂S·9H₂O; purity >98%, ACS reagent; ACROS Organics; New Jersey; USA), sodium sulphite (NaSO₃ anhydrous; purity >98%, ACS; Bernd Kraft der Standard; Duisburg; Germany) and sulphuric acid (H₂SO₄; SRM3154; NIST; USA) were used to investigate the effect of different sulphur species on the separation. Three different ion exchange resins were used in this study. The details of these resins are shown in Table 2. All resins were activated and cleaned before use.

Table 2 Information on ion exchange resins

a Depends on the amount of copper (2 mL for copper ≤20 mg).
Property	AmberliteCG50 (ref. 14)	AG1X8 (ref. 15)	Chelex-100 (ref. 16)
Company	Sigma-Aldrich	Biorad labs	Biorad labs
Resin type	Weak cation exchange	Strong anion exchange	Weak cation exchange
Functional group	Carboxylic acid	Quaternary ammonium	Carboxylic acid
Ionic form	H^+	Cl^−	Na^+
Size (mesh)	100–200	200–400	200–400
Total exchange capacity (mmol mL^−1)	3.5	1.2	0.4
Selective to copper	High	None	Very high
Function in the separation procedure	Removes copper	Retains sulphur on the resin	Removes copper
Amount of resin (mL)	2^a	1	1

---

The copper reference materials BAM-M385, BAM-M376a, BAM-228 and BAM-227, all produced by BAM, were selected to serve as well-defined samples for the development of the sulphur–matrix separation procedure (for details see Table 1).

2.2 Equipment and instrumentation

The samples were weighed by using an analytical balance (Mettler Toledo AX205, Giessen, Germany). When plastic labware was used for weighing, the labware was flushed with a nitrogen ion stream to remove electrostatic charges. Digestion was accomplished using a high pressure asher, HPA (Anton Paar GmbH Graz, Austria) equipped with a heating block holding 5 quartz digestion vessels of 90 mL volume. The digestion program lasted 4 h at a maximum temperature of 320 °C and a maximum pressure of about 130 bar. The temperature was increased from room temperature to 150 °C in 30 min and then up to 320 °C in 60 min. This temperature was kept for 150 min followed by a cooling step down to room temperature with a duration of 60 min.

All mass spectrometric measurements were performed using a sector field ICP-MS instrument element 2 (Thermo-Fisher Scientific, Germany), unless stated otherwise. Table 3 shows the operating conditions of the instrument.

Table 3 Instrument operating parameters for sulphur measurements

Parameter	Setting
Instrument type	Element 2
Autosampler	Cetac ASX 100
Aspiration mode	Self-aspirating
Nebulizer	MicroMist 100 μL
Spray chamber	Cyclonic spray chamber
Interface	Jet interface
Cones	Ni sampler and skimmer X-cone
Cool gas flow rate	16 L min^−1
Auxiliary gas flow rate	0.8–1.0 L min^−1
Sample gas flow rate	0.9–1.25 L min^−1
RF power	1200 W
Guard electrode	On
Mass resolution mode	Medium
Acquisition mode	Pulse and analog mode
Runs/passes	10/40
Sensitivity in cps/(μg g^−1)	1 × 10^7 for ^32S
Drift correction	Yes

---

2.3. Sample preparation

Sulphur standards. Secondary stock solutions were prepared from the parent solutions of NIST SRM3154 and the ³⁴S spike solution. They were kept separately stored under controlled conditions and they were monitored for their weight before and after each withdrawal to enable correction of evaporation loss. From these solutions working standards were diluted in 2% HNO₃ gravimetrically.

Dissolution and digestion/oxidation/equilibration of copper samples. Copper metal was processed to yield small pieces, 0.1 g to 0.25 g of which were weighed into HPA vessels. In the case of IDMS, the spike ³⁴S was added before sample dissolution aiming at a ³²S/³⁴S ratio of 1. Then 5 mL of conc. HNO₃ was slowly added. This step must be carried out carefully and the work was performed in a fume hood due to the strong reaction between the metal and conc. HNO₃ producing large amounts of toxic gas (NO_x). After this step, it was required to wait until the copper was completely dissolved. Then 1 mL of H₂O₂ was added before digestion which was carried out by applying the HPA. The digested solution which contained copper depending on the sample type was weighed again before it was split into subsamples.

Sulphur–matrix separation procedure. Das et al. published an improved sulphur separation procedure for the determination of δ³⁴S values in standards and in seawater by MC-ICPMS.¹⁷ Their method required less processing time and less consumption of chemicals, and it offered lower procedure blanks (12–250 ng of sulphur) compared to previously published methods. The recovery calculated for sulphur standard solutions reached 100 ± 2%. For low or simple matrix containing samples the separation procedure of Das et al. is well suited. Therefore, we applied it to the quantification of sulphur in biodiesel by IDMS (see the Method validation section). For the separation of sulphur from a copper matrix this procedure does not work, and a completely new procedure had to be developed.

The complete sulphur–copper separation process is visualized in Fig. 1. The newly developed procedure for sulphur–copper separation consists of three subsequent separation steps and is described in the following in detail. Approximately 1.0–1.5 g of the digested sample solution (light blue colour, b) was weighed into Savillex PFA beakers and was evaporated to dryness at 110 °C. The residue then was dissolved in 8 mL of Milli-Q water followed by the addition of conc. ammonia in excess at room temperature (at least >10 times the amount of ammonia as calculated on the basis of the chemical reaction per mole of copper, c). The colour of the sample solution turned a characteristic deep blue due to the formation of the copper–ammonia complex. To this solution CG50 resin (white) was added in excess (1 mL: 10 mg Cu, d), and the resulting suspension was mixed well by using an automatic shaker for 3 h. The resin turned light blue whereas the solution was clear and transparent. In the meantime, 1 mL of anion exchange resin (AG-1X8) was packed in a column (2 mL PP + PE column, I.D. 0.7 cm, e) and rinsed with Milli-Q water. The column was closed at the lower end with parafilm and was then loaded with the clear solution from above. After about 20 min the parafilm was removed to let the remaining matrix pass through. The remaining CG50 resin from above was rinsed with 4 mL Milli-Q water which was then loaded onto the AG 1X8 column (this was repeated 4 times). Thereafter, the sulphur fraction was eluted from the AG1X8 resin by addition of 12 mL of 0.25 M HNO₃ onto the column. The eluted solution was evaporated to dryness at 110 °C overnight. Typically, the residue was blue or green (f) which means that remains from the copper matrix are still present. Therefore, the Chelex resin was used to remove the rest of the copper. Approximately 1 mL of Chelex resin was packed into the column (g), which was closed at the lower end with parafilm. In parallel, the residue was dissolved in 2 mL of Milli-Q water and was loaded onto the column. After 20 minutes, the parafilm was removed and the sulphur was eluted with 10 mL of Milli-Q water. Then the sulphur containing solution was evaporated to dryness at 110 °C. Finally, the residue was dissolved in 2% HNO₃ (h) such that a final sulphur mass fraction of approximately 2 μg g⁻¹ was obtained, based on the known sulphur mass fraction in the solid sample and the recovery. This was verified within ±20% for each sample by comparing to a 2 μg g⁻¹ sulphur standard.

Fig. 1 Sulphur–copper separation process; (a) Cu sample, (b) digested sample solution, (c) copper–ammonia complex solution, (d) sample solution after adding weak action (CG50) resin, (e) separation with anion exchange (AG1X8) resin, (f) residue after separation by anion exchange, (g) separation with Chelex resin, (h) residue and final solution, and (i) ICP-MS measurement.

Isotope dilution mass spectrometry (IDMS). The IDMS calibration approach is based on the addition of an isotopically enriched spike of the analyte element to an unknown amount of the analyte element to be determined in the sample. In fact, IDMS might also be regarded as an internal standard method which uses isotopes of the analyte element itself as an ideal internal standard and measures the ratio of the main isotope (analyte, ³²S) and the spike isotope (spike, ³⁴S) and then compares the ratio before spiking with those obtained after spiking. The large difference and major advantage of IDMS is that the analyte elements in the spike and sample behave approximately similarly, which is not the case in normal internal standardization. The equation (eqn (1)) used to quantify the mass fraction of sulphur in this work is displayed below. Details on the deduction, calculation, relation to other IDMS equations and correction factors are given in Vogl et al.¹⁰where w_x = mass fraction of sulphur in the sample (μg g⁻¹), w_y,b = mass fraction of sulphur in the spike (μg g⁻¹), M_x = standard atomic weight of sulphur in the sample, M_b = atomic weight of the spike isotope, a_x,b = abundance of the spike isotope in the sample, m_x = mass of the sample (g), m_y = mass of the spike (g), R_x = isotope ratio in the sample, R_y = isotope ratio in the spike, R_xy = isotope ratio in the sample blend, and ProcBlank = procedure blank (μg)

For each IDMS analysis, the ICP-MS measurement sequence started with the sulphur standard (also named back-spike), then the unspiked sample, sample blend, procedure blank and at the end the spike, with regularly measuring a standard or back-spike in between.

3. Results

Whenever uncertainties are provided they are expanded measurement uncertainties with k = 2 calculated from combined uncertainties via U = k × u_c, unless stated otherwise.

Sulphur measurement

The major challenges of sulphur measurement by ICP-MS are high background, high first ionization potential (10.36 eV),¹⁸ light mass causing high mass discrimination, and spectral interference at m/z 32 and 34. The main interferents are ¹⁶O¹⁶O⁺, ³¹P¹H⁺, ¹⁴N¹⁸O⁺, ¹⁵N¹⁶O¹H⁺, and ⁶⁴Zn²⁺ on ³²S⁺ and ¹⁶O¹⁸O⁺, ³²S¹H¹H⁺, ¹⁶O¹⁶O¹H¹H⁺, and ⁶⁸Zn²⁺ on ³⁴S⁺. To separate all interferents the medium resolution mode (m/Δm > 4300) was required. The intensities of ⁶⁴Zn⁺ and ⁶⁸Zn⁺ were monitored (3 × 10⁵ and 2 × 10⁵ cps for ⁶⁴Zn and ⁶⁸Zn, respectively, <10 ng g⁻¹) to observe any upcoming problems with high intensities of doubly charged Zn isotopes. ⁶³Cu⁺ and ⁶⁵Cu⁺ were used to investigate the matrix removal efficiency. The corresponding Cu signal intensities were 0.04–2 × 10⁷ cps for ⁶³Cu and 0.02–1.5 × 10⁷ cps for ⁶⁵Cu (<150 ng g⁻¹) in the final solution.

The signal intensities observed for ³²S in 2% HNO₃ (instrumental background) were in the range of 0.2–1.3 × 10⁶ cps; for procedure blanks they were 0.4–2.5 × 10⁶ cps and for sulphur standards at 2 μg g⁻¹ they were 2–3 × 10⁷ cps. As a consequence, the standard signal for 2 μg g⁻¹ was 20 times above those of the blank in all cases, except the standard addition IDMS for NIST SRM494, where it was only 10 times above the blank level. The relative standard deviation of the sulphur isotope ratio over all back spike measurements within one sequence was around 1.3% relative. This number expressed not only the reproducibility of the ratio measurement, but also indicated instrumental drift and wash-out effects during the whole IDMS sequence.

S-Species vs. digestion/oxidation/equilibration conditions

Most copper is produced by mining and/or extraction from copper sulphide and thus sulphur impurities might also occur in the form of sulphides. Unfortunately, the anion exchange resin (AG1X8) is selective to sulphate and sulphite but less-selective to sulphide.¹⁵ Therefore, when quantifying total sulphur in copper, the different species of sulphur need to be oxidized to sulphate prior to the sulphur–matrix separation on the AG1X8 resin in order to avoid any measurement bias. Consequently, the recovery of the different sulphur species was checked for the applied separation procedure. Stock solutions of sulphate, sulphite and sulphide were gravimetrically prepared from sulphuric acid, sodium sulphite and sodium sulphide. Then portions of the species standards containing about 10 μg sulphur were oxidized separately in two different ways: (1) addition of 3 mol L⁻¹ HNO₃ and H₂O₂ (30%, w/w) and hotplate oxidation at 120 °C for 3 h; (2) addition of 3 mol L⁻¹ HNO₃ and H₂O₂ (30%, w/w) and digestion with HPA. After this oxidation, the samples were subjected to the AG1X8 separation procedure and subsequently spiked. The samples were measured by ICP-IDMS; the sulphur amount was calculated and subsequently the recovery for each sample was calculated referring to the gravimetrically determined input. The results are displayed in Table 4. For sulphate, of course quantitative recovery was obtained. In the case of sulphite, the hotplate oxidation is insufficient (79% recovery), whereas the HPA oxidation yields quantitative recovery (100%). In the case of sulphide, either oxidation experiment is insufficient yielding recoveries below 50%. Therefore, another oxidation experiment was carried out applying the HPA oxidation as described above but with concentrated nitric acid (65%, w/w) followed by AG1X8 separation, evaporation to dryness and redissolution in 2% nitric acid. This time, the samples were measured with ICP-MS applying external calibration. The obtained recovery for sulphide was 94% with an estimated expanded uncertainty of 10%. Thus, the oxidation of sulphide to sulphate can be considered complete with a quantitative recovery for the AG1X8 separation procedure.

Table 4 Oxidation conditions for the conversion of different sulphur species into sulphate and the corresponding recovery rates after separation by using the AG1X8 resin measured by ICP-IDMS or ICP-MS together with expanded uncertainties (recovery related to the gravimetric value)

Species	% recovery
	3 M HNO3 + H2O2 by hot plate (120 °C, 3 h)^a	3 M HNO3 + H2O2 by HPA^a	Conc. HNO3 + H2O2 by HPA^b
a IDMS applied. b % recovery from the external calibration method.
Sulphide, S^−	15	37	94 ± 10
Sulphite, SO3^−	79	100 ± 1	n.a.
Sulphate, SO4^2−	100 ± 1	100 ± 1	n.a.

---

When applying the HPA oxidation with concentrated HNO₃ and H₂O₂ a complete conversion from sulphide and sulphite to sulphate could be achieved. The recovery of all investigated sulphur species is quantitative within measurement uncertainties.

Sulphur–matrix separation

The performance of the separation procedure was evaluated mainly by checking the sulphur recovery and the efficiency of matrix removal.

As a starting point the separation procedure used by Das et al.¹⁷ was applied. This separation procedure employs a strong anion exchange resin (AG1X8), which retains the sulphur on the column while the matrix elutes without retardation. The recovery of this procedure was checked with a sulphur standard solution (sulphate form) and was found to be 100 ± 2%. Then the procedure was applied to synthetic sample solutions (sulphur 8 μg g⁻¹ and copper 24 000 μg g⁻¹). The recovery of sulphur dropped to 10–30%. An explanation for the low recovery could be the formation of copper(II) sulphate (CuSO₄·(H₂O)_x) complexes by the reaction of sulphate with excess copper. Therefore, the complexation of copper by ammonia being a highly selective ligand for Cu(II) was explored. The formation of the tetraamine-copper(II) complex releases trapped sulphate and leads to increased recovery rates. The chemical reaction of ammonia and copper is well known and is shown in eqn (2) and (3). Upon ammonia addition, the sulphur recovery for the synthetic sample increased to 100 ± 3 (n = 4).

CuSO₄(aq) + 2NH₃(aq) + 2H₂O(l) → Cu(OH)₂(s) + (NH₄)₂SO₄(aq) (2)

Cu(OH)₂(s) + 4NH₃(aq) → [Cu(NH₃)₄](OH)₂(aq) (3)

This procedure was then applied to real world samples (solid copper metal). The recovery of sulphur dropped to 10–20% again. Possible reasons for this considerable decrease are the very high amount of the copper matrix compared to the sulphur mass fraction (approximately 3 times higher than in the synthetic sample) and the occurrence of different sulphur species in the sample. It was assumed that the removal of the major part of the copper prior to the AG1X8 separation should solve this problem. Unfortunately, most cation exchange resins being capable of separating copper are strong acidic cation resins containing sulphonated polystyrene as the exchange site. This however would lead to unacceptably high procedure blanks. When considering weakly acidic cation resins a suitable material could be identified: the resin CG50 does not contain sulphur groups and is capable of retaining copper.^19,20 Together with its high capacity (3.5 mmol mL⁻¹) this makes it highly suitable for the intended task. The new separation step was carried out by adding an excess amount of cation resin CG50 (1 mL: 20 mg Cu) into the sample solution. The deep blue solution turned into a clear and transparent solution, while the resin itself turned from white to blue. The clear solution was loaded onto the column which contains AG1X8 to further purify the sulphur fraction. After eluting the sulphur from the AG1X8 resin the solution was dried on a hot plate until dryness yielding a blue/green residue which still contains copper above 100 μg g⁻¹. Therefore, another chelating ion exchange resin (chelex100) was employed to trap the remaining copper.²¹

The copper samples investigated in this study contain copper in the range of 0.85–0.99 kg kg⁻¹ and zinc from <10 to 300 g kg⁻¹. Approximately 0.10–0.25 g of these samples were used to perform the sulphur–copper separation. After applying the complete three stage separation procedure the mass fractions of both elements were significantly reduced to below 400 ng g⁻¹ for copper and below 50 ng g⁻¹ for zinc, respectively. Nearly the complete matrix (>99.999%) was removed which results in an extremely high matrix removal factor of above 10⁵.

Measurement results of sulphur in copper by ICP-IDMS

The above described sulphur–matrix separation was applied and combined with the described ICP-IDMS procedure for the sulphur quantification in six different reference materials. The results are displayed in Table 5. Samples BAM-M385, BAM-M376a, BAM-228, BAM-227 and SRM494 were quantified by normal IDMS whereas SRM494 and SRM1034 were quantified by a combination of the standard addition technique and the IDMS technique because the content of sulphur was lower than the working range of the separation procedure. The obtained measurement results for samples BAM-M376a, BAM-228 and SRM494 agree well with the reference values (Table 1) whereas for BAM-M385 and BAM-227 the results differ from the certified values. Sample BAM-227 was certified in 1979 by photometrical methods only. At that time, metrological concepts such as measurement uncertainty and SI traceability were not clear or even not existing, and therefore the reference value was expressed in terms of precision only. In the case of sample BAM-M385 the IDMS result obtained within this work and the certified value are significantly different. The uncertainty of the certified value also represents only the dispersion of the inter-laboratory comparison.

Table 5 Sulphur mass fractions in copper reference materials as obtained by ICP-IDMS and individual uncertainty contributions

List		BAM-M385	BAM-M376a	BAM-228	BAM-227	NIST SRM494	NIST SRM494^a	NIST SRM1034^a
a Combined standard addition and IDMS technique. b For the type of reference refer to Table 1.
Measurement value and MU (μg g^−1), k = 2		(37.72 ± 0.19)	(133.68 ± 0.86)	(385.50 ± 2.40)	(1376.60 ± 6.2)	(14.34 ± 0.09)	(14.97 ± 0.20)	(6.79 ± 0.36)
Relative measurement uncertainty (%)		0.5	0.6	0.6	0.5	0.7	1.34	5.30
Reference value^b		(31.2 ± 1.5)	(133 ± 19)	(360 ± 40)	(1220 ± 100)	(15 ± 3)	(15 ± 3)	(2.8 ± 0.2)
Cu mass fraction in the final solution (ng g^−1)		<100	<150	<100	<150	<150	<260	<370
Zn mass fraction in the final solution (ng g^−1)		<10	<50	<10	<10	<10	<10	<10
Number of replicates		8	8	8	8	4	6	6
Uncertainty budget	Type	% contribution
Observed ratio of the back spike	A	48.1	58.5	65.5	33.7	41.8	11.1	2.8
Mass fraction of the spike	B	42.4	27.3	28.3	55.0	24.3	45.1	81.2
Observed ratio of sample blends	A	5.2	8.2	2.4	4.0	29.7	22.4	2.4
Observed ratio of natural	A	2.7	4.8	1.2	1.1	1.6	0.1	<0.1
Observed ratio of the spike	A	<0.1	<0.1	0.8	0.4	1.8	2.0	<0.1
Weighing of samples	A	0.4	<0.1	1.2	4.7	<0.1	<0.1	<0.1
Weighing of spikes	A	<0.1	<0.1	<0.1	<0.1	<0.1	5.0	2.3
Procedure blank	B	<0.1	<0.1	<0.1	<0.1	<0.1	<0.1	<0.1
Mass fraction of the back spike (standard addition)	B	—	—	—	—	—	13.2	10.7
Weighing of the back spike (standard addition)	A	—	—	—	—	—	<0.01	<0.01
Others	—	1.2	1.2	0.6	1.1	0.8	1.2	0.6

---

The quantification of sulphur in NIST SRM494 by conventional IDMS is hindered by the very high Cu/S ratio, which clearly affects the separation in a negative way: the recovery of sulphur dropped to about 30% for four replicates, while two further replicates even showed recoveries below 10%, revealing a non-linear recovery function with a threshold amount of sulphur which is lost during the whole separation procedure. The resulting samples show sulphur amounts too low for reliable ICP-MS measurements. To enable measurement without completely changing the separation procedure, an exact amount of sulphur was added prior to spiking, such that the sulphur mass fraction was shifted to the optimum working range of the separation procedure. So, exact amounts of sulphur were added prior to spiking to enhance the mass fraction of sulphur from 15 μg g⁻¹ to 40 μg g⁻¹, and then the IDMS analysis was performed as usual and finally the added sulphur amount was subtracted. The so obtained measurement result agreed well with the certified value within the uncertainties. The measurement uncertainties for the IDMS experiment with the addition of a known amount of sulphur were approximately twofold those obtained with the IDMS experiment without addition of a known amount of sulphur.

The same concept of the addition of a standard solution was applied to NIST SRM1034 by increasing the sulphur content from 3 μg g⁻¹ to 40 μg g⁻¹. However, in this case the measurement result was significantly different from the certified value. It has to be noted here that NIST SRM 1034 was certified by a single method only in 1982. This disagreement requires further investigation. The measurement results for NIST SRM494, however, proved that the standard addition technique combined with IDMS is a suitable tool to extend the working range of the separation procedure. This combination provides reliable results which are true within the stated uncertainties as shown for NIST SRM494.

The developed procedure for the quantification of low sulphur amounts in copper has been validated here via three different routes, which will be explained in the following chapter: first an inter-laboratory comparison at the highest metrological level (see the Method validation section), second a step-by-step validation by checking each single step of the procedure and third the setup of a complete uncertainty budget. Additionally, SI traceability is provided in the most direct way. These metrological properties make the IDMS results more likely to be the true values for the materials BAM-M385 and BAM-227 rather than the certified values.

Method validation

Quantification of the sulphur mass fraction in biodiesel fuel by ICP-IDMS. One approach for validation is the participation in inter-laboratory comparisons. This approach was used to verify that the developed IDMS procedure is capable of accurately determining sulphur mass fractions at the low μg g⁻¹ level. The only available comparison for low sulphur measurements during the research period was CCQM-K123 “trace elements in biodiesel fuel”. Although the matrix did not fit with a copper matrix and thus the complete sulphur–copper separation was not directly tested, the comparison was at the highest metrological level and could at least validate the spiking, the ³⁴S spike solution, the digestion step, the anion exchange step (AG1X8) and the isotope ratio measurement as well as the calculations. Furthermore, it could verify the level of the obtained measurement uncertainty (corresponding to formerly used terms of accuracy and precision). The biodiesel fuel sample was digested by using the HPA;¹³ the sulphur was isolated by anion exchange chromatography (AG1X8 resin) and then measured by using a MC-ICP-MS (Neptune Plus, Thermo Scientific, Bremen) in combination with an Aridus desolvation system and the addition of sodium in a weight ratio of 2 : 1 (Na : S). The so obtained result shows an excellent agreement with the other results and demonstrates that the developed procedure enables sulphur measurements at the low μg g⁻¹ level in complex matrices with sufficiently low measurement uncertainties (Fig. 2).²²

Fig. 2 Results of the CCQM-K123 inter-laboratory comparison: the mass fraction of sulphur in biodiesel fuel displayed for the participating laboratories together with the reference value (all error bars represent expanded uncertainties, k = 2) based on the data from ref. 22. BAM's result was (7.39 ± 0.10) μg g⁻¹ while the reference value was (7.38 ± 0.35) μg g⁻¹.²²

Detection limit. The procedure blank was determined by IDMS as well and yielded procedure blanks for the individual IDMS measurement sequences ranging from 3 ng to 53 ng. The average of these individual procedure blanks (n = 22) was calculated and yielded a total procedure blank of 14 ng sulphur with a standard deviation of 12 ng. The limit of detection (LOD, blank + 3SD) calculated on this basis was 0.20 μg g⁻¹ while the limit of quantification (LOQ, blank + 10SD) was 0.54 μg g⁻¹, when considering a sample weight of 0.25 g.

However, LOD and LOQ are more of a theoretical concept for reference measurements applying IDMS, because the applicability of IDMS procedures is more strongly defined by the working range, which itself is limited by the separation procedure and the measurement uncertainty aimed at. When applying the same separation procedure without any adaption and when aiming at relative measurement uncertainties of <2%, a working range from approximately 15 μg g⁻¹ to 1500 μg g⁻¹ could be established. For samples containing sulphur mass fractions of <15 μg g⁻¹ the addition of an accurately weighed amount of standard is necessary, as explained above for the sample NIST SRM494.

Measurement uncertainty. Measurement uncertainty is an indirect term enabling a quantitative definition of the previously used terms accuracy and precision of a measurement value. The measurement uncertainty gives the dispersion of measurement results related to the performance of the measuring system, which includes the true value and which also covers the precision term.

Within this study ICP-IDMS was applied as a higher-order reference measurement procedure or in other terms a primary ratio method of measurement, where the measurement process is well understood and a measurement equation can be written down, permitting the calculation of the mass fraction of sulphur directly from the signal intensities. Consequently, measurement uncertainties were assessed based on the IDMS equation. The individual contributions to the measurement uncertainty listed for each sample type are displayed in Table 5. The main contribution to the uncertainty accounting to >30% derived from the observed isotope ratio in the back spike for conventional IDMS; this is caused by the relative standard deviation of the sulphur isotope ratio of the back spike which amounts to approximately 1.3% for a complete measurement sequence as mentioned before. The second largest contribution is made up by the mass fraction in the spike (>24%), followed by the observed isotope ratio in the sample blend, the unspiked sample and the spike (<10%). All other quantities do not contribute significantly (<2%). This also applies for the very low procedure blank (average value 14 ng). In the case of the modified IDMS, where back-spike is added to the sample before spiking, the main contribution to the measurement uncertainty is made up by the mass fraction of the spike and the back spike.

The relative expanded measurement uncertainties for conventional IDMS are below 1%. When applying the modified IDMS, where back-spike is added to the sample before spiking, the relative expanded measurement uncertainties are larger and amount to 1.34% and 5.30% as calculated for sample no. SRM494 and SRM1034, respectively.

Metrological traceability. Metrological traceability is defined as the “property of a measurement result whereby the result can be related to a reference through a documented unbroken chain of calibrations, each contributing to the measurement uncertainty”.^23,24 When establishing an unbroken chain of calibrations and thus SI traceability, the measurement result is considered reliable, acceptable and comparable. The metrological traceability to the SI for the mass fraction of sulphur, w_x, in copper metal is established by such an unbroken chain of comparisons, each accompanied by an uncertainty budget.

The establishment of the metrological traceability of the sulphur content in copper samples is visualized in Fig. 3 for the example of sample no. BAM-M376a. The metrological traceability chain for double IDMS shows the course leading from the mol and/or kg down to the final sulphur mass fraction in the sample.

Fig. 3 Metrological traceability chain of IDMS measurement results for the sulphur content in sample no. BAM-M376a.

The boxes on the left-hand side express measurement results with measurement uncertainties of the calibrator or sample, while the boxes on the right-hand side show details of the measuring systems and the measurement procedures. The arrows in the middle of the chart express the action of calibrators and measuring systems.

From the bottom of the calibration hierarchy, the three boxes (from left to right) describe the measurement uncertainty, the measurement result, the sulphur mass fraction and the sample, here BAM-M376a. The measurement result was assigned by applying the IDMS approach at BAM (labelled measurement procedure 4 in Fig. 3) as described in this research. For the IDMS approach, the exact mass fraction of the ³⁴S spike solution is required (second calibrator 2) which was obtained by reverse IDMS which is labelled measurement procedure 3 in Fig. 3. The reverse IDMS was applied by using the primary assay NIST SRM3154 as back-spike. NIST SRM3154 was certified by two measuring systems which were gravimetric and coulometric titration at NIST (the information is displayed in the certificate) and represents the primary calibrator 1 for the sulphur mass fraction. The sulphur mass fraction of the primary calibrator 1 is in turn metrologically traceable to the definition of the SI measurement unit mole through the quantity values for electric current and kilogram. This makes the sulphur mass fraction of sample BAM-M376a, as obtained by the here described IDMS procedure, traceable to the SI in the most direct way.

4. Conclusions

In this study, a new analytical procedure for the sample preparation and quantification of the sulphur mass fraction in copper and copper alloys by ICP-IDMS was developed. This includes the purification of sulphur by a three-stage separation procedure consisting of the complexation of copper with ammonia, cation exchange and anion exchange chromatography. The procedure features a high performance which is expressed in a matrix removal efficiency of above 99.999% while the overall sulphur recovery stays above 80%, which is more than sufficient for IDMS applications. Procedure blanks are in the range between the maximum values of 4 ng and 53 ng, with the majority of the blank values ranging from 9 ng to 16 ng. This is 7 times lower than those reported for ID-TIMS¹⁴ whereas LOD and LOQ values are 0.20 μg g⁻¹ and 0.54 μg g⁻¹, respectively being the same level as previous work but it has to be noticed that the matrices are different. The procedure blank obtained in this study contributed to the combined uncertainty below 0.1% whereas in ID-TIMS the contribution is 36–99%.¹³ As mentioned before, the LOD and LOQ are more theoretical concepts for IDMS including trace-matrix separation and have only limited practical use, because the working range of the analytical procedure is defined by the matrix separation procedure and the measurement uncertainty aimed at. With the presented sulphur–matrix procedure a working range from approximately 15 μg g⁻¹ to 1500 μg g⁻¹ can be achieved. In addition to the low detection limits, the here presented procedure offers a high matrix removal efficiency, low measurement uncertainties and SI traceability.

The presented analytical procedure was successfully validated via three different routes, first a partial validation via an inter-laboratory comparison at the highest metrological level,²² second a step-by-step validation of the whole analytical procedure, and third the setup of a complete uncertainty budget. Additionally, most of the certified values of the analysed reference materials agree well with the results obtained by the IDMS procedure. Expanded relative measurement uncertainties were estimated to range below 1% while metrological traceability to the SI is clearly expressed. Therefore, the procedure is well suited to provide reference values for the total sulphur mass fraction in copper materials.

Thus, the procedure reported in this study is established as a reference procedure for sulphur measurement in copper being fit for the following purposes: certification of reference materials, assignment of reference values for inter-laboratory comparison and calibration of routine analytical methods such as GDMS, LA-ICPMS, X-ray fluorescence, and carrier gas hot extraction (CGHE) via matrix-matched standards calibrated by the here described analytical procedure.

Conflicts of interest

There are no conflicts to declare.

List of abbreviations

BAM	Bundesanstalt für Materialforschung und -prüfung (BAM)
BCR	Community Bureau of Reference
BE	Belgium
Cps	Count per second
CCQM	Consultative Committee for Amount of Substance
CENAM	The National Metrology Center, Mexico
CGHE	Carrier gas hot extraction
CRM	Certified reference material
Cu	Copper
DE	Germany
GDMS	Glow discharge mass spectrometry
HPA	High pressure asher
ICP-MS	Inductively coupled plasma-mass spectrometry
IDMS	Isotope dilution mass spectrometry
INMETRO	National Institute of Metrology, Quality and Technology, Brazil
K	Coverage factor
LA-ICPMS	Laser ablation inductively coupled plasma-mass spectrometry
MU	Measurement uncertainty
NIMT	National Institute of Metrology (Thailand)
NIST	National Institute of Standards and Technology
NMIJ	National Metrology Institute of Japan
RM	Reference material
S	Sulphur
SI units	International System of Units
SRM	Standard reference material
TIMS	Thermal ionization mass spectrometry
UME	National Metrology Institute, Turkey
UK	United Kingdom
US	United States
VIM	International Vocabulary of Metrology

Acknowledgements

The authors thank D. Becker and M. Koenig for assistance and support with the lab work. Especially, the first author thanks the Federal Institute for Materials Research and Testing (Bundesanstalt für Materialforschung und -prüfung (BAM), Germany), the Office of the Civil Service Commission (OCSC, Thailand) and the National Institute of Metrology (NIMT, Thailand) for financial support.

Notes and references

1. Copper Development Association (CDA), http://copperalliance.org.uk/industry/economy, accessed, 19 Feb, 2017.
2. R. Matschat, J. Hinrichs and H. Kipphardt, Anal. Bioanal. Chem., 2006, 386, 125–141.
3. L. Li and R. W. Messler, Weld. J., 1999, 78, 387s–396s.
4. Y. S. Sayi, P. S. Shankaran, C. S. Yadav and G. C. Chhapru, Indian J. Chem. Technol., 2003, 10, 373–381.
5. Laboratory of the government chemist (LGC), Copper reference materials, 2016.
6. J. G. Martinez-Sierra, O. G. San Bias, J. M. M. Gayon and J. I. G. Alonso, Spectrochim. Acta, Part B, 2015, 108, 35–52.
7. M. C. R. Matschat, M. Hamester and S. Pattberg, Fresenius' J. Anal. Chem., 1997, 359, 418–423.
8. B. Lange, S. Recknagel, M. Czerwensky, R. Matschat, M. Michaelis, B. Peplinski and U. Panne, Microchim. Acta, 2008, 160, 97–107.
9. M. Sargent, R. Harte and C. Harrington, Guidelines for Achieving High Accuracy in Isotope Dilution Mass Spectrometry (IDMS), Royal Society of Chemistry, 2002.
10. J. Vogl and W. Pritzkow, Mapan-J Metrol Soc I, 2010, 25, 135–164.
11. W. R. Kelly and P. J. Paulsen, Talanta, 1984, 31, 1063–1068.
12. W. R. Kelly, L. T. Chen, J. W. Gramlich and K. E. Hehn, Analyst, 1990, 115, 1019–1024.
13. W. Pritzkow, J. Vogl, R. Koppen and A. Ostermann, Int. J. Mass Spectrom., 2005, 242, 309–318.
14. Rohm and Haas Company, Specification's product, 2007.
15. Bio-Rad Laboratories, Instruction Manual: Strong Anion Exchange Resin AG 1, AG MP-1 and AG 2.
16. Bio-Rad Laboratories, Instruction Manual: Chelex® 100 and Chelex 20 Chelating Ion Exchange Resin.
17. A. Das, C. H. Chung, C. F. You and M. L. Shen, J. Anal. At. Spectrom., 2012, 27, 2088–2093.
18. https://en.wikipedia.org/wiki/Ionization_energy, accessed, 24 June, 2017.
19. R. Biesuz, M. Pesavento, G. Alberti and F. Dalla Riva, Talanta, 2001, 55, 541–550.
20. M. Pesavento and E. Baldini, Anal. Chim. Acta, 1999, 389, 59–68.
21. M. Pesavento, R. Biesuz, A. Profumo and T. Soldi, Environ. Sci. Pollut. Res. Int., 2003, 10, 317–320.
22. T. Kuroiwa, Y. Zhu, K. Inagaki, S. Long, S. Christopher, M. Puelles, M. Borinsky, N. Hatamleh, J. Murby, J. Merrick, I. White, D. Saxby, R. C. de Sena, M. D. de Almeida, J. Vogl, P. Phukphatthanachai, W. Fung, H. Yau, T. O. Okumu, J. N. Kangiri, J. A. Salas Téllez, E. Z. Campos, E. C. Galván, N. Kaewkhomdee, S. Taebunpakul, U. Thiengmanee, C. Yafa, N. Tokman, M. Tunç and S. Z. Can, Metrologia, 2017, 54(Tech. Suppl.), 08008.
23. JCGM200:2012, JCGM, 3rd edn, 2012, ch. 2008, version with minor corrections.
24. P. De Bievre, R. Dybkaer, A. Fajgelj and D. Brynn Hibbert, Pure Appl. Chem., 2011, 83, 1873–1935.

---